Social Rocket
Agencies are increasingly supplementing their existing client contracts with AI-specific clauses instead of rewriting master service agreements from scratch. Following conversations between Digiday and four agency executives, the focus is primarily on data use, intellectual property, liability, tool disclosure and human oversight.
The development is more than detailed legal work: generative and agentic AI is simultaneously changing how agencies produce services, staff teams and bill clients. This puts contract logic and business models under adaptation pressure at the same time.
AI enters first as an addendum to the MSA
The agencies Digiday spoke with currently describe a predominantly pragmatic approach. Instead of drafting new contract sets for each client relationship, existing MSAs are supplemented with addenda or individual clauses.
According to its own account, Mile Marker regulates, among other things, how metadata is handled, who holds rights to AI-generated content and which tools are used. Go Fish has added further rules on brand safety and data protection over the past six months. Crispin is also gradually supplementing its contracts.
The pattern is not yet an industry standard – Digiday's sample covers only a few companies. But it concretely shows which questions are now regularly coming up on the client side.
Five points are becoming central for marketing teams
For clients, five particularly practical contract questions can be derived from this development:
- Which AI tools may be used for the account?
- Which client, campaign or personal data may enter these systems?
- Who holds which rights to inputs and outputs?
- What human review is required before publication or delivery?
- Who bears the risk for rights infringements or faulty outputs?
The last point becomes especially complicated when systems do not merely suggest content but carry out actions themselves. An analysis by Mayer Brown therefore describes, for agentic AI, a shift away from pure SaaS contracts toward models with clearer service definitions, audit rights, liability rules and outcome-based service levels.
Compliance comes into play in the EU
For German agencies and brands, the topic has become even more concrete since 2August 2026. Since that date, the transparency obligations under Article 50 of the EU AI Act have applied. The European Commission requires, among other things, transparency in certain AI interactions as well as labelling or machine-readable marking for certain AI-generated or manipulated content.
This does not mean that every ad edited internally with AI must automatically be visibly labelled. Contract clauses should, however, not only address copyright and data protection but also clarify who checks whether a specific use falls under a legal transparency obligation and who documents the implementation.
The Commission points out that violations of the relevant transparency rules can be penalised with fines of up to 15 million euros or 3 percent of worldwide annual turnover – depending on the circumstances and taking proportionality into account.
The bigger problem is remuneration
AI saves production time. That is exactly what clashes with the traditional agency model, in which many services are calculated based on hours and team capacity.
Digiday therefore describes a second contractual issue: if the same work is done with fewer human hours, pure time-based billing loses its logic. Some agencies are developing proprietary AI tools or services, but do not yet know whether these should be billed as software, an additional service, a licence or part of an outcome-based price.
This fits the wider transformation of the industry. WPP is aligning its agency model more closely with AI and new remuneration models, while at Omnicom, data, creativity and AI are moving closer together following the IPG integration.
A report summarised by MediaPost 4As-/Forrester with almost 200 surveyed US agency decision-makers also shows the economic uncertainty: 61 percent described AI as ongoing operating costs, and only 6 percent as a separate revenue line. This underlines why contracts are currently changing faster in terms of risk and governance than in the actual pricing logic.
An AI addendum is already worthwhile for clients
Marketing teams do not have to wait until a new industry standard becomes established. A clean AI addendum can already today record permitted tools, data classes, review processes, IP rules, liability and disclosure obligations, while also providing for regular reviews.
This becomes especially important as more and more operational work is automated. At Meta Ads, AI-generated creatives and automatic delivery logic are already growing directly into the platform. The boundary between "the agency uses a tool" and "the system carries out work independently" is therefore becoming increasingly blurred.
The real change 2026 is therefore not a single new contract clause: AI forces agencies and clients to renegotiate rights, data, control and payment at the same time.